The study applies and expands the routine activity theory to examine the dynamics of online harassment and violence against women on Twitter in India. We collected 931,363 public tweets (original posts and replies) over a period of 1 month that mentioned at least one of 101 influential women in India. By undertaking both manual and automated text analysis of "hateful" tweets, we identified three broad types of violence experienced by women of influence on Twitter: dismissive insults, ethnoreligious slurs, and gendered sexual harassment. The analysis also revealed different types of individually motivated offenders: "news junkies," "Bollywood fanatics," and "lone-wolves", who do not characteristically engage in direct targeted attacks against a single person. Finally, we question the effectiveness of Twitter's form of "guardianship" against online violence against women, as we found that a year after our initial data collection in 2017, only 22% of hostile posts with explicit forms of harassment have been deleted. We conclude that in the social media age, online and offline public spheres overlap and intertwine, requiring improved regulatory approaches, policies, and moderation tools of "capable" guardianship that empower women to actively participate in public life.